Which models run on 20 GB of VRAM?

Largest model that runs comfortably: Qwen3.5 9B

• Qwen3.5 0.8B1 GB packageComfortable
• Qwen3.5 2B2.7 GB packageComfortable
• Qwen3.5 4B3.4 GB packageComfortable
• Qwen3.5 9B6.6 GB packageComfortable
• gpt-oss-20b14 GB package · the maker states a 16 GB memory minimumRuns
• Qwen3.5 27B17 GB packageRuns
• Qwen3.5 35B24 GB packageLimited
• gpt-oss-120b65 GB package · the maker states a 80 GB memory minimumNot suitable
• Qwen3.5 122B81 GB packageNot suitable

• ComfortableRoom for the model file and a long context
• RunsFits; the context window may need to stay short
• LimitedNot enough memory; part of the model spills to system RAM and runs slowly
• Not suitableNot a GPU-run model on this card

Rule: 1.5× the requirement in memory is Comfortable, 1.15× Runs, 0.6× Limited; the requirement is the default package size, or the maker's stated minimum where there is one (a stated minimum counts as fitting at exactly that figure). Package sizes are the Ollama library's, minimums the maker's own statement; read on 2026-09-11. Sources: ollama.com/library/qwen3.5, ollama.com/library/gpt-oss, openai.com/index/introducing-gpt-oss

All cards and models